Graphics Reference
In-Depth Information
3.2
Local Tone Mapping
After the global adaptation processing, the local tone mapping based on retinex theory
is applied to enhance the local contrast and preserve the details. The use of Gaussian
function as surround function in retinex algorithm will cause the halo artifacts along
high-contrast edges. But the halo artifacts can be reduced by introducing an
edge-preserving filter, so
L
0
smoothing filter is introduced as surround function in our
algorithm. The local tone mapping can be expressed as the following formula:
(8)
Lxy
(,
)L gF((,
=
Lxy
)L gF((,
−
Wxy
)
out
g
in which
L
is the result of local adaptation;
W
is the
L
0
-smoothed version of
L
;
out
and LogF( ) represents a logarithm function as the follow formula:
log(100
L
+
ʵ
ʵ
)
LogF(
L
)
=
(8)
log(100
+
)
where
ʵ
refers to the nonlinearity offset. The logarithm function is a nonlinear
function whose gradient is gradually decreasing. A small
ʵ
can effectively enhance the
contrast of dark areas.
3.3
Color Correction
After the local adaptation, the processed luminance values are rescaled from 0 to 1.
Finally, the tone mapped image is obtained from the processed luminance and the
original RGB image. Similar to other approaches [7], the color correction is processed
as the following formula:
s
R
in
L
out
L
w
R
s
out
G
G
=
in
L
(8)
out
out
L
w
B
out
s
B
in
L
out
L
w
in which,
B
represent the tone mapped RGB image.
L
is the luminance of original image, and
ou
L
is the luminance after local tone mapping. The exponent
s
controls the color
saturation, which is usually set between 0.4 and 0.6. The result is then linearly scaled
to the range of 0-255 for visualization.
R
,
G
and
B
represent the original RGB image;
R
,
G
and
in
in
in
out
out
out
Search WWH ::
Custom Search